Mutations in AML: prognostic and therapeutic implications

Acute myeloid leukemia (AML) accounts for ∼25% of all adult-onset leukemias in the Western world, with an incidence of 3 to 5 cases out of every 100 000 adults per year and a median age of 69 years at diagnosis.^1-3  AML is a hematopoietic stem cell (HSC) malignancy characterized by differentiation arrest and uncontrolled clonal proliferation of neoplastic immature precursors, preventing normal bone marrow hematopoiesis. The prognosis of AML is variable, based in part on clinical features such as patient age, medical comorbidities, and performance status, and also on underlying genetic features including both cytogenetic and molecular aberrations.^4,5 

Treatment of AML has changed little in over 40 years; curative therapy remains the nucleoside analog cytarabine, administered in combination with an anthracycline as induction therapy, followed with repeated cycles of high-dose cytarabine and/or an allogenic stem cell transplant (SCT) in patients attaining a complete remission (CR).⁶  The administration of intensive cytarabine-based therapy is appropriate regardless of age in “medically fit” older patients⁷ ; hypomethylating agent (HMA) strategies are also often recommended and used as frontline treatment approaches in elderly patients.^8,9  A recent analysis of treatment patterns and outcomes in elderly AML patients by the Surveillance, Epidemiology and End Results-Medicare database demonstrated survival benefit with the receipt of either intensive or HMA strategies, compared with the 60% of elderly AML patients in the United States who received no leukemia-directed therapy.¹⁰  Current outcomes with these standard AML approaches remain highly unsatisfactory; patients ≤60 years of age demonstrate a long-term disease-free survival (DFS) of only ∼40%, whereas DFS in older patients is a sobering 10% or less, with a median overall survival (OS) less than 1 year regardless of therapeutic approach.^11-13  These poor outcomes with traditional therapies, particularly when contrasted with the effectiveness of targeted therapy such as all-trans retinoic acid and arsenic trioxide for the AML subset acute promyelocytic leukemia (M3 AML),^14,15  offers hope that the improved understanding of AML leukemogenesis will lead to personalized and genomically targeted therapeutic approaches, which will improve outcomes in additional AML subsets.

Cytogenetic analysis to detect large structural chromosomal abnormalities provided the first “genetic” prognostication schema in AML, and remains the backbone of current AML genomic classification, partitioning patients based on their pretreatment karyotype into those with favorable, intermediate, or adverse cytogenetics, and correlating with 5-year OS of ∼60%, 30% to 40%, and 5% to 10%, respectively.^4,16-18  With the advancement of next-generation sequencing approaches, recurrent mutations in AML have also been characterized, and utilizing current technology >95% of AML cases are confirmed to harbor at least 1 somatic alteration, with approximately a dozen genomic alterations identified per AML sample including an average of 3 driver mutations, and providing particularly useful prognostic information in the majority of AML patients with otherwise intermediate-risk cytogenetics.^19-21  The genetic profile of AML is notably heterogeneous, and only a few mutations (eg, Fms-related tyrosine kinase 3 [FLT3], nucleophosmin [NPM1], and DNA methyltransferase 3A [DNMT3A]) are present in more than a quarter of AML patients,²⁰  reinforcing the understanding that the clinical diagnosis of AML encompasses a diverse group of genetically distinct malignancies.

The focus of this review is to summarize the genomic landscape of AML with a highlight on somatic mutations, underscoring the utility of a comprehensive genomic assessment. Attention to recent clinical advances, including novel molecularly targeted therapies for treatment of AML, will also be highlighted.

At the time of AML diagnosis, prognostically important and distinct chromosomal translocations and large chromosome gains or losses can be observed on standard karyotype analysis in approximately half of AML patients.⁴  Over the past decade, DNA sequencing technology has revealed recurrent AML gene mutations, undetectable by standard cytogenetic analysis, which additionally contributes to AML pathogenesis. Many of these gene mutations have prognostic implications, independently and/or in the presence of certain co-occurring driver mutations (gene-gene interactions), and targeted therapeutic strategies directed at specific mutations and molecular classes are currently under development. Although the following list is not meant to be exhaustive, a summary of the most frequent recurrent genetic abnormalities in AML are described in the sections to follow, and detailed in Table 1. Of note, the presence of mutations within a functional category is largely mutually exclusive, suggesting functional overlap between recurrent mutations within functional groups. Moreover, the individual prognostic information derived from the presence or absence of a specific mutation can be modified by the presence of cooperating co-mutations (ie, NPM1 with FLT3-internal tandem duplications [ITDs] or DNMT3A mutations), signifying awareness of the complete AML genomic landscape will be necessary to provide optimal personalized prognostication.

Present in approximately two-thirds of AML cases, mutations leading to aberrant activation and proliferation of cellular signaling pathways, frequently referred to as type 1 mutations, make up the most common mutational subset in AML. Intriguingly, mutations within this class are frequently identified in subclonal cellular fractions, indicating they are often late clonal events in disease evolution.²¹ 

ITDs and/or activating kinase domain point mutations (D835) in the FLT3 gene are present in nearly one-third of patients with AML, resulting in constitutive activation and downstream signaling through the RAS/RAF/MEK/mammalian target of rapamycin growth and proliferation pathways, and the phosphatidylinositol 3 kinase (PI3K)/AKT prosurvival pathway. FLT3-ITD mutations are more frequent within younger adult patients, those with normal karyotype (NK), and are associated with proliferative AML (eg, higher white blood cell and blast %), and increased risk of relapse leading to decreased OS.²²  In addition to the presence/absence of FLT3 mutations, the degree of mutational burden as estimated by the FLT3 variant allele frequency (VAF) is of importance, with AML patients with low FLT3-ITD burden (often considered as a VAF <0.5), especially in the setting of concomitant NPM1 mutation experiencing relatively improved outcomes,²³  suggesting that both the FLT3-mutational status and FLT3-VAF should be reported in all AML patients for optimal prognostication.

The family of RAS oncogenes is mutated in ∼10% to 15% of AML cases. These include activating mutations in NRAS, KRAS, PTPN11, and NF1, leading to aberrant proliferative signaling through the RAS/RAF/MEK/extracellular signal-regulated kinase pathway. Although the prognostic impact of RAS mutations has been debated, the acquisition or clonal outgrowth of RAS mutations, analogously to FLT3, can be seen at the time of myelodysplastic syndrome (MDS) progression to AML and portends a poor outcome in this circumstance.²⁴  The effect of NRAS mutations (specifically those within codon G12/13) was recently described to confer a favorable outcome in the setting of an NPM1 and DNMT3A co-mutated genotype, with minimal prognostic importance in other molecular permutations.²¹  Development of directed therapeutic targets for RAS-mutant leukemias, such as MEK, extracellular signal-regulated kinase, PI3K, and mammalian target of rapamycin inhibitors are ongoing, but have been complicated by the rapid development of resistance or activation of bypass pathways in the single-agent setting.^25,26 

Nearly exclusively identified in AML with core-binding factor (CBF) translocations, mutations in the receptor tyrosine kinase KIT, particularly the D816V missense variant, occur in ∼20% of CBF-AML and are frequently associated with an adverse prognosis, although this has not been universally observed.^27,28  The addition of multikinase inhibitors with activity against KIT mutations, such as midostaurin or dasatinib, may be of benefit in this otherwise favorable cohort. In updated results of a phase 2 trial of 7 + 3 chemotherapy in combination with dasatinib 100 mg orally daily (days 8 to 21) for newly diagnosed CBF-AML patients, clinical outcomes were comparable to historical CBF-AML cohorts (90% CR rate and 65% 2-year DFS), with no difference in outcomes between KIT-mutated and wild-type (WT) patients, confirming the safety of this approach.²⁹ 

The first generation of “FLT3” inhibitors incorporates several multikinase inhibitors (eg, sorafenib, sunitinib, midostaurin, and lestaurtinib) employed for the treatment of FLT3-mutated AML. The lack of FLT3 selectivity with these multikinase, or “dirty kinase” first-generation inhibitors, has been frequently (and perhaps naively) viewed to represent a suboptimal FLT3-targeted therapeutic strategy, and much effort has been devoted to the development of more selective second-generation small molecule FLT3 inhibitors such as quizartinib (AC220), crenolanib, and gilteritinib (ASP-2215).

A double-blind, placebo-controlled, phase 2 trial of adults aged 18 to 60 years with newly diagnosed AML, randomized to 7 + 3 induction therapy vs 7 + 3 + sorafenib 400 mg twice daily was recently reported.³⁰  A total of 267 patients were enrolled, and a significant difference in event-free survival (EFS) with a median EFS of 9 months vs 21 months and a 3-year EFS of 22% vs 40% was demonstrated. A total of 46 enrolled patients (17%) had FLT3-ITD mutations; the EFS benefit was seen regardless of FLT3-ITD mutational status, suggesting that inhibition of kinases other than FLT3 may have provided antileukemic activity in this newly diagnosed, younger AML cohort. Whether this study indicates that sorafenib should be added to the induction regimen of all younger AML patients, regardless of molecular status, is an ongoing debate.

The landmark results of the phase 3 international prospective RATIFY study of adults (n = 717) aged 18 to 60 years with newly diagnosed FLT3-mutated AML, treated with 7 + 3 + midostaurin vs placebo, were recently reported.³¹  The 5-year EFS rate in the midostaurin vs placebo arm of 26.7 vs 19.1 (P = .004), and 5-year OS of 50.8 vs 43.1 (P = .007), respectively, confirms that pretreatment genetic assessment improves treatment decisions and patient outcome. In February 2016, the US Food and Drug Administration granted midostaurin “breakthrough therapy” designation status for patients with newly diagnosed, FLT3-mutated AML. A global compassionate use program and US Expanded Treatment Protocol are currently in place to enable midostuarin access. Midostaurin inhibits multiple kinases in addition to FLT3, including CKIT, platelet-derived growth factor receptor, vascular endothelial growth factor receptor, and protein kinase C. Thus, whether midostaurin will provide antileukemia activity in FLT3 WT but otherwise class I addicted AML is a valuable question to be evaluated in future trials.

Several phase 2 studies have been completed in the relapsed/refractory AML setting, with encouraging single-agent activity including a composite CR rate (CRc) of 44% to 54% and modest median response duration of 11 to 12.7 weeks, due primarily to the development of resistance mutations within the FLT3 tyrosine kinase domain (eg, D835), or gatekeeper mutations (eg, F691) when used in the single-agent setting.^32,33  A phase 3 trial of quizartinib monotherapy vs salvage chemotherapy in relapsed/refractory FLT3-ITD mutant AML is ongoing (#NCT02039726).

Initially developed as a platelet-derived growth factor receptor inhibitor, crenolanib demonstrates pan-selective FLT3 activity, active against both the FLT3-ITD and FLT3-D835 resistance mutation. Several phase 2 studies of crenolanib in combination with induction chemotherapy, or as post-SCT maintenance, are ongoing (#NCT02283177 and #NCT02400255).

Gilteritinib has potent activity against both FLT3-ITD and D835 mutations, and also the gatekeeper mutation F691L. Interim results from the phase 1/2 trial of relapsed/refractory AML were reported last year, with a CRc rate of 43% as a single-agent therapy. Notably different than the first generation of multikinase inhibitors, AML patients with WT FLT3 status receiving gilteritinib derived nominal benefit, with a CRc <10%. A randomized phase 3 trial of gilteritinib vs investigator choice salvage chemotherapy is ongoing for relapsed/refractory FLT3-mutant AML (#NCT02421939), and a randomized phase 3 trial (#NCT02752035) of gilteritinib alone, vs gilteritinib + azacitidine, vs azacitidine alone, in FLT3-mutant newly diagnosed AML, began recruitment in June 2016.

Somatic mutations within key epigenetic regulators are identified in >50% of AML, and are now recognized as a key, and often an inciting, component of leukemogenesis.²¹  It is of interest that age-related clonal hematopoiesis, identified in >10% of individuals over age 65, is predominantly defined by the clonal outgrowth of preleukemic clones harboring mutations in one of the genes within this epigenetic class.^34,35  It is likely that mutations in this class of epigenetic modifiers promote clonal outgrowth, but are insufficient to initiate leukemic transformation without subsequent mutational events.

One of the 3 most common mutations in AML, DNMT3A, is recurrently mutated in ∼20% of de novo AML. Operating via de novo methylation of cytosines in cytosine guanine dinucleotide islands, DNMT3A plays a critical role in regulation via epigenetic silencing of HSC differentiation and self-renewal. The most frequent DNMT3A mutation, the R882 missense mutation, has been demonstrated to prevent methyltransferase activity and DNA binding, leading to impaired HSC function, increased self-renewal, and a differentiation block.^36,37 DNMT3A mutations frequently occur with advanced age and in conjunction with NPM1, FLT3-ITD, and/or isocitrate dehydrogenase 1 (IDH1) mutations; the prognostic importance of DNMT3A mutations by themselves remains unclear, although several series have shown a correlation with resistance to chemotherapy and disease relapse.³⁸ 

Inactivating mutations throughout TET2 occur in 10% to 20% of MDS and AML, with a prominent frequency up to 50% in patients with chronic myelomonocytic leukemia. The resultant decrease in enzymatic TET2 function in the setting of mutations impedes the breakdown of 5-methylcytosine to 5-hydroxymethylcytosine, leading to increased HSC self-renewal, impaired myeloid differentiation, and a hyper-methylated epigenetic signature.³⁹  The prognosis of patients with TET2 mutations is variable and likely dependent on the presence of additional cooperating pathogenic events.

Loss-of-function mutations in the chromatin modifier gene ASXL1 occurs in 10% to 20% of AML and are associated with advanced age, antecedent hematologic malignancy, concurrent runt-related transcription factor-1 (RUNX1) and spliceosome mutations, and poor outcomes.⁴⁰  ASXL1 is a chromatin-binding protein that interacts with the polycomb repressor complex 2, including the H3K27 methyltransferase EZH2; mutant ASXL1 decreases polycomb repressor complex 2 recruitment and stabilization, and reduces repressive H3K27 trimethylation marks at various loci.⁴¹ 

The IDH family of metabolic enzymes operates within the Krebs cycle, catalyzing the oxidative decarboxylation of isocitrate to α-ketoglutarate (α-KG). Recurrent “hot spot” mutations within the IDH enzymatic active sites (IDH1-R132, IDH2-R140, and IDH2-R172) occur in ∼20% of AML, and are more frequent in intermediate-risk AML, advanced age, and concurrent with NPM1 mutations.⁴²  IDH1/IDH2 mutations lead to an alternate reaction which reduces α-KG into the oncometabolite 2-hydroxyglutarate, which then competitively inhibits α-KG–dependent reactions, including TET2-dependent DNA hydroxymethylation, histone demethylase activity, B-cell lymphoma 2 (BCL2)-dependent mitochondrial respiration, and the activation of the hypoxia-inducible factor-1α pathway via prolyl hydrolase inhibition.⁴³  Of interest, mutations in IDH1, IDH2, and TET2 appear to be nearly mutually exclusive, suggestive of a unique epigenetic mutational class leading to a global hypermethylation signature. IDH2-R172 mutations in particular appear to result in a unique gene expression profile with more profound metabolic aberrations, and are suggested to identify a distinct subgroup of 1% of AML patients with fewer co-mutations and a favorable prognosis with intensive chemotherapy.²¹ 

Targeted small molecule inhibitors of IDH2 (enasidenib, formerly AG221), IDH1 (AG120, IDH305, and FT-2102), or pan-IDH1/IDH2 (AG881) are currently in clinical development, with encouraging early results. Interim results of the phase 1 study of single-agent enasidenib in refractory hematologic malignancies show an overall response rate (ORR) (defined as CR/CR with incomplete blood count recovery [CRi]/morphologic leukemic-free state/partial remission) of 41% in 198 patients, with 18% CR rate in a primarily relapsed/refractory AML population.^44,45  Interval results of the phase 1 study of AG120 (#NCT02074839) describes similar efficacy in a smaller (n = 65) patient population, with an ORR of 36% and a CR rate of 18%.⁴⁶  Several additional clinical trials with AG120 and enasidenib are ongoing, including 7 + 3 + enasidenib or AG120 (#NCT02632708) and azacitidine + enasidenib or AG120 (#NCT02677922), both as frontline therapy for newly diagnosed AML populations, and a randomized phase 3 “IDHENTIFY” trial of enasidenib vs investigator’s choice for relapsed/refractory AML with IDH2 mutation (#NCT02577406). Preliminary results of the phase 1 (#NCT02381886) study of the IDH1-inhibitor IDH305 are anticipated by the end of 2016.

Small molecule inhibitors of bromodomain-containing “reader” proteins, which recognize acetylated lysine residues and direct chromatin remodeling, include the bromodomain and extraterminal domain family and bromodomain inhibitors OXT015, CPI-0610, TEN-010, GSK525762, and INCB054329. These inhibitors have demonstrated formidable preclinical activity, and several are currently under evaluation in phase 1 AML studies (#NCT02158858, #NCT02308761, #NCT01943851, and #NCT02308761).

In view of the epigenetic instability introduced by mutations in epigenetic modifier genes, the HMAs such as azacitidine, decitabine, and guadecitabine (SGI-110) are frequently recommended for the treatment of AML patients harboring these mutations, particularly given the more advanced age of AML patients with epigenetic mutations. HMAs inhibit DNA methyltransferase-1 and are known to induce hypomethylation, although the mechanisms responsible for antileukemia activity remain poorly understood and there has been no confirmed relationship between the presence or absence of epigenetic mutations, degree of demethylation with HMA treatment, and patient outcome.^47,48  Of interest among MDS patients, the combination of mutant TET2 with WT ASXL1 demonstrated an improved HMA response rate (60% vs 43%) although without OS impact,⁴⁹  further indicating that evaluation of gene-gene interactions, and not only single mutations, will be essential to better identify treatment strategies and provide improved prognostication.

Along with FLT3 and DNMT3A, NPM1 is one of the 3 most frequent driver mutations in AML. NPM1 mutations, occurring almost exclusively within exon 12 of the gene, occur in approximately one-third of adults with AML, and in more than 50% of NK-AML, frequently in the context of mutations in epigenetic modifiers such as DNMT3A, TET2, IDH1, or IDH2 mutations. The NPM1 gene encodes for the nuclear chaperone protein NPM, which shuttles between the nucleus and cytoplasm and plays a role in diverse cellular functions, including protein formation, ribosome biogenesis, DNA replication, and the cell cycle. NPM1 mutations are typically stable throughout the disease course, and are identified in nearly all leukemic cells and impart a distinct expression profile.⁵⁰ NPM1-mutant AML, specifically those without FLT3-ITD mutations, consistently demonstrates augmented chemosensitivity to standard therapy and overall favorable outcomes, even among older patients.^51,52  However, NPM1 mutations in the setting of mutant DNMT3A, particularly in the setting of FLT3-ITD, confer a markedly poor prognosis.^21,53  Some reports suggest distinctive activity of all-trans retinoic acid and actinomycin D in AML with mutant NPM1, although this observation has not been systematically confirmed.^54,55 

Approximately 20% to 25% of adults with AML have mutations involving the myeloid transcription factors RUNX1, CCAAT/enhancer binding protein α (CEBPA), and/or GATA2. To date, no targeted or otherwise specific therapeutic options exist for this prognostically disparate molecular subgroup.

CEBPA is a key hematopoietic transcription factor involved in lineage-specific myeloid differentiation. Mutations in the single exon CEBPA gene occur in ∼10% of AML, predominantly in younger patients and NK, or otherwise intermediate-risk karyotypes. Pathogenic CEBPA mutations occur within 2 discrete gene regions: N-terminal domain frameshifts directing an alternate 30-kDa protein, and C-terminal domain insertions and deletions, which disrupt DNA-binding and dimerization. Originally included in the 2008 World Health Organization (WHO) AML classification system, AML with mutated CEBPA has been confirmed as a unique pathologic entity within the 2016 revised classification schema, with the important distinction that biallelic mutations (encompassing 2/3 of CEBPA-mutant patients) are associated with this favorable prognosis.^56-58 

RUNX1 is a critical transcription factor, essential during embryogenesis in HSC generation, and during adulthood in the regulation of HSC differentiation and homeostasis. Various missense and frameshift mutations in RUNX1 are identified in AML patients at a frequency of ∼10% to 15%, and are particularly frequent in patients with antecedent hematologic disorders or secondary AML, advanced age, increased cytopenias at diagnosis, shorter OS, and chemoresistance to standard therapy.⁵⁹  Mutations in RUNX1 are thought to establish a compromised stem cell phenotype characterized by early HSC exhaustion.⁶⁰  The 2016 revised WHO AML classification system has added AML with mutated RUNX1 as a provisional entity, owing to the biologically distinct molecular signature and poor prognosis with this molecular AML subtype.^59,61 

The GATA2 gene encodes a zinc-finger transcription factor with a major role in myeloid hematopoiesis, via cooperation within a network of transcription factors in a tightly regulated and dose-dependent manner. Assorted gene-wide mutations resulting in GATA2 haploinsufficiency include truncating, missense, and noncoding variants within the regulatory GATA2 region.⁶² GATA2-mutated AML represents ∼5% of all patients, although notably GATA2 mutations are seen as a cooperating event in ∼25% of CEBPA-mutated AML patients and do not appear to impact the favorable outcome of dmCEBPA.⁶³ 

Although mutations in myeloid malignancies are nearly always presumed to be somatic events, recent evidence suggests that a portion of hematologic cancers, similar to solid tumors, have an underlying hereditary cancer predisposition syndrome.⁶⁴  Because germ line samples in patients with hematologic malignancies are not typically obtained, standard sequencing technology is unable to distinguish whether a mutation is germ line or somatic in nature. The germ line inheritance of mutations in the transcription factor genes CEBPA, RUNX1, and GATA2 lead to well-described, autosomal dominant, familial predispositions to AML, which are notably now outlined within the 2016 WHO AML guidelines.⁵⁶  Attention to the medical and family history, and consideration of germ line testing must be considered when mutations in the above genes are identified.⁶⁵ 

Approximately 10% of patients with biallelic CEBPA mutations are predicted to have an N-terminal germ line mutation, with a secondary somatic C-terminal CEBPA mutation identified at the time of AML diagnosis. The favorable prognosis of biallelic CEBPA mutations is not affected by whether both mutations are somatic or one is of germ line inheritance.⁶⁶  Notably, there are no prodromal signs, symptoms, or antecedent cytopenias described with this syndrome that would help augment clinical suspicion. Individuals with germ line RUNX1 mutations are diagnosed with the syndrome “familial platelet disorder with propensity to myeloid malignancies,” which is characterized by variable degrees of life-long thrombocytopenia and aspirin-like platelet dysfunction. The long-term outcome of “familial platelet disorder with propensity to myeloid malignancies” is variable, with an ∼35% lifetime risk of hematologic malignancy at an average age at diagnosis of 33 years.⁶⁷  In addition to increased lifetime risk of MDS/AML, germ line GATA2 mutations are further characterized by dramatic immunodeficiency related to profound B lymphocyte, NK, and dendritic cell deficiencies, frequent atypical and viral infections, and other nonhematologic symptoms including pulmonary, rheumatologic, lymphatic, and vascular pathology comprised within the “MonoMAC Syndrome.”⁶⁸ 

Frequently mutated in MDS and myeloproliferative neoplasms, mutations in splicing factors (SF3B1, SRSF2, U2AF1, and ZRSR2) are identified in ∼10% of patients with AML and are associated with older age, less proliferative disease, poor rates of response to standard treatment, and decreased survival. Spliceosome mutations are postulated to promote malignancy through missplicing of various genes involved in epigenetic regulation, transcription, and genome integrity.⁶⁹  Current evidence posits spliceosome mutations are mutually exclusive of each other and always co-expressed with the WT allele, suggesting that one functional allele is required for cellular integrity. Thus, a synthetically lethal approach of splicing inhibitor therapy for patients with preexisting splicing factor mutations has been postulated to be a uniquely effective targeted therapy for this genetically defined AML subset.⁷⁰  Preclinical and phase 1 studies are ongoing, including a phase 1 trial of the selective small molecule splicing modulator (H3B-8800), specifically for patients with hematologic malignancies (#NCT02841540).

Termed “the guardian of the genome,” TP53 is a key tumor suppressor with protean functions related to maintenance of genomic stability, including regulation of cellular senescence, apoptosis, metabolism, and DNA repair. Although uncommon in de novo AML, TP53 mutations occur in ∼15% of therapy-related AML or AML with MDS-related changes, and are predominantly associated with complex cytogenetics, advanced age, chemoresistance, and poor survival.^71,72  Irrespective of age or treatment modality, TP53 mutations in AML portend lower response rates and inferior outcomes compared with TP53 WT AML patients,⁷³  including high relapse rates after SCT. The optimal treatment strategy for TP53-mutant patients thus remains a critical area of unmet need.

With activity extending outside of a specific molecular group, BCL2 inhibitors, along with other inhibitors of antiapoptotic proteins such as MCL1, are promising small molecule approaches in AML therapy (Table 2). Despite modest single-agent activity in the relapsed/refractory AML setting with an ORR of 19%, the BCL2 inhibitor venetoclax (formerly ABT-199) has demonstrated impressive early responses in the frontline treatment of elderly AML patients unfit for standard chemotherapy, with a CR/CRi rate of 71% and median time to CR/CRi of 29.5 days.⁷⁴  Elegant preclinical work has demonstrated striking synthetic lethality of the IDH-mutant AML to BCL2 inhibition owing to the aberrant oncometabolite 2-hydroxyglutarate elevation in IDH mutants, rendering these leukemic cells particularly dependent on BCL2.⁷⁵  Clinical responses have encouragingly been seen across AML subgroups, and an analysis of BCL2 protein sensitivity index at presentation may provide a useful predictive biomarker for response.⁷⁶ 

The MDM2 protein, which functions in complex with TP53 and is frequently overexpressed in TP53 WT AML, can be inhibited by small molecule MDM2 inhibitors, leading to TP53 stabilization and restoration of TP53 activity. The MDM2 inhibitor RG7112 demonstrated modest single-agent activity in TP53 WT patients, with baseline MDM2 expression levels predictive of response.⁷⁷  The efficacy and tolerability of novel MDM2 inhibitors, particularly in synergistic combinations, is keenly awaited.

Patients with AML have traditionally been characterized within 1 of 3 risk groups based on the presence of various cytogenetic abnormalities, with the majority of patients historically falling within the NK or otherwise intermediate-risk group. For the first time, in 2010, recurrent somatic mutations, specifically mutations in NPM1, FLT3, and CEBPα, were incorporated to further refine prognostic information.¹⁸  It is now estimated that two-thirds of the variation in AML outcome is based on genomic features, including fusion genes, copy-number alterations, point mutations, and gene-gene interactions.²¹  The current 2016 revision to the WHO AML classification reinforces the favorable prognostic importance of AML with mutated NPM1 (without concurrent FLT3 mutation), and AML with biallelic CEBPA mutations.⁵⁶  Similar to patients with CBF AML, patients with NPM1-mutant and dmCEBPA AML appear to be particularly chemosensitive to anthracycline and cytarabine-based combination chemotherapy approaches, and allogeneic SCT is not recommended at the time of first CR.^50,78  Thus, in addition to CBF AML patients, patients with NPM1-mutant or dmCEBPA AML, regardless of age, should receive intensive induction and consolidation therapy when possible. Furthermore, the addition of FLT3-inhibitor therapy at diagnosis for patients with FLT3-ITD–mutated AML has demonstrated significant survival benefit and this should now be considered the standard of care. New to the 2016 WHO classification, AML with mutated RUNX1 has been incorporated as a provisional entity, due to the apparent biologically distinct RUNX1-mutated AML subgroup, demonstrating both a poor response to standard cytotoxic therapy and poor overall outcomes.⁵⁹  The best treatment algorithm for RUNX1-mutated patients, as well as patients with other high-risk genetic lesions such as chromatin-spliceosome mutations, TP53 mutations, or complex cytogenetics has not been established, and well-designed clinical trials remain the most appropriate option.

A unique AML classification schema has been described among patients with secondary or therapy-related AML; secondary defining AML arising from an antecedent hematologic disorder (secondary AML) and therapy-related AML (tAML) characterizing patients having previously received leukemogenic therapies for nonmyeloid disorders. The presence of mutations within 8 genes (ASXL1, SRSF2, SF3B1, U2AF1, ZRSR2, EZH2, BCOR, and STAG2) demonstrated >95% sensitivity in defining secondary AML compared with de novo AML, notably all genes commonly mutated in MDS and involved in chromatin modification, cohesion complex, and the spliceosome.⁷¹  In patients with tAML, the presence of TP53 mutations in 15% was associated with a highly complex karyotype, chemoresistance, and poor survival. A separate cohort of tAML patients was identified who were genetically classified as “de novo AML,” with the presence of genetic events such as NPM1 mutations and CBF rearrangements; these patients had outcomes consistent with de novo AML, including favorable response to standard cytotoxic chemotherapy and improved EFS. These results advocate that the evaluation of genetic “ontogeny” in AML may provide more prognostic and therapeutic information than the traditional classification.

The importance of the identification of recurrent mutations in AML extends beyond the initial treatment decision. Evaluation of somatic mutations can be followed over time, especially as reproducible and standardized methods of molecular minimal residual disease (MRD) assessment monitoring continue to improve (see Ossenkoppele, in this book⁷⁹ ). This is particularly important in the cohort of NK-AML patients, for whom cytogenetic and fluorescence in situ hybridization methods of MRD analysis are not routinely available. For example, a “favorable risk” NK-AML patient in CR1 with NPM1-mutant, FLT3-WT status, yet with persistent mutant NPM1 detection by MRD assessment after cycle 2 of therapy, is associated with a higher risk of relapse, and an SCT in CR1 would be indicated based on real-time molecular MRD results, despite the favorable original prognostic score.⁸⁰ 

In relapsed/refractory AML, few standard treatment options exist, and patient outcome remains poor. An enhanced understanding of the patient-specific mutational AML landscape at relapse can inform salvage treatment programs, and repeat cytogenetic and molecular classification at the time of relapse is warranted to determine whether patients are eligible and/or appropriate for investigational or alternative treatment options. Patients with favorable-risk AML may benefit most from re-induction cytarabine-based strategies. An evaluation for FLT3, IDH1, and IDH2 mutations should be universal at relapse, given the promising results with the respective targeted inhibitors available in late-phase clinical trials. In cytarabine-refractory AML patients with poor-risk molecular or cytogenetic features, clinical trial options incorporating hypomethylating-agent combinations, novel targeted therapeutics, or other treatment modalities (ie, antibody-based and/or immunotherapy approaches) are appropriate strategies.

Given the extraordinary recent progress made in the understanding of cancer biology and mechanisms of leukemogenesis, we are now uniquely poised to see our research-based innovations translate directly into the care of patients with AML. Encouraging single-agent efficacy has been demonstrated with targeted small molecule inhibitors such as FLT3, IDH1, and IDH2, and the early evidence of synergistic activity of the BCL2 inhibitor venetoclax in combination with HMA, speaks to the importance of rational AML combinations. The ultimate realization of this recent sequencing era will be to organize the diverse genomic information into a distinct framework to inform our prognostic accuracy and provide personalized treatment decisions for all.

Courtney D. DiNardo, Department of Leukemia, University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Unit 0428, Houston, TX 77030; e-mail: cdinardo@mdanderson.org.